การจับคู่สไลซ์ระหว่างภาพ CT และ MRI บริเวณเนื้องอกอุ้งเชิงกราน โดยใช้โครงสร้างของกระดูก

Abstract

This engineering project aims to develop a slice matching method between Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) in patients with pelvic tumors using bone structures as anatomical references for image similarity comparison. Since CT and MRI differ in the number of slices, slice thickness, and image positions, direct slice-to-slice matching is challenging. This study used medical images from 10 patients. Bone structures were segmented from CT images using threshold segmentation and from MRI images using a U-Net model. The images were then rescaled and aligned before slice matching. Three slice matching methods were compared: matching using Dice Similarity Coefficient (DSC), matching using a combined Similarity Score based on DSC, Area Ratio, and Bounding Box Ratio, and the proposed method using Bone Component Similarity Matching. Experimental results from 145 reference slice pairs showed that the proposed method achieved the best performance, with a Mean Absolute Error (MAE) of 9.24, Accuracy of 13.10%, Accuracy within ±1 slice of 35.17%, and Accuracy within ±2 slices of 51.72%. The results demonstrate that considering bone component-level similarity can improve slice matching accuracy and can be used as a preprocessing step for CT–MRI image registration.